Abstract
Purpose :
Ophthalmic examination in large population-based cohorts is challenging and time-consuming. To address this, (1) we built a myopia proxy that can be applied in large population-based studies and (2) validated the proxy by confirming its association with educational attainment and a polygenic risk score (PRS) for myopia.
Methods :
Data were collected between 2014-2017 from 88,646 Dutch adults from the Lifelines cohort. We did a three-step analysis. In Step 1, we performed principal component analysis (PCA) to responses of five refraction-status questions posed to Lifelines participants. In Step 2, we measured the refractive state in a subset of Lifelines participants (n=209) and performed logistic regression using myopia (mean spherical equivalent [MSE] < -0.5 D) as dependent and the principal components (PCs) (with an eigenvalue ≥ 1) as independent variables. We used ROC curve analyses to identify specificity, sensitivity, and the classification threshold. In Step 3, the classification equation was applied to the remaining Lifelines participants from Step 1. The value of the proxy was then explored by determining the prevalence of myopia and calculating its association with educational attainment and a PRS of myopia. Analyses were adjusted for age and sex.
Results :
77,213 participants (58.1% females) were eligible for the PCA. Of these, about three-quarters (73.4%) were wearing either glasses or contact lenses. Using measured MSE as a gold standard, the first two PCs had a specificity (95% CI) of 92.4% (87.3-96.6%), and sensitivity (95% CI) of 89.1% (80.0-96.4%) for myopia. The area under the ROC curve (95% CI) was 94.5% (90.6-98.4%). The age-standardized prevalence (95% CI) of myopia was 34.7% (34.3-35.1%) . Compared to low education level, the odds ratios (95% CI) of myopia were 1.60 (1.53-1.70, P= 4.04x10-82) and 2.40 (2.30-2.54, P =2.19x10-252) for medium and high education levels, respectively (Figure A). Similarly, PRS analysis showed a dose-response relationship with myopia (Figure B). Compared to the lowest decile of the PRS, the highest decile had an OR (95% CI) of 4.24 (3.68-4.91, P=5.27x10-86) for myopia.
Conclusions :
Self-administered refractive error questions are effective in capturing myopic cases in a population-based setting.
This is a 2021 ARVO Annual Meeting abstract.